Baidu unveils ERNIE Evolve, an automated research system that beat public baselines on GPU kernel and training tasks
Baidu_Inc · x · 2026-11-09
Baidu introduced ERNIE Evolve, an automated research system for real-world AI R&D, positioned as a first step toward recursive self-improvement.
How it works: given a research goal, code repo, optimization objective and evaluation criteria, the system autonomously picks research directions, modifies code, and runs experiments in isolated environments. Only improvements passing independent evaluation and review are kept; each round's findings guide the next.
Results:
- Beat baselines on three public tasks: NanoChat AutoResearch (training algorithms), NanoGPT Speedrun (training speed), and SOL-ExecBench (GPU kernel optimization), outperforming the strongest public results on the first two
- Applied to inference optimization for the in-development ERNIE-5.5-Flash, with changes passing production-grade validation
Signals that major labs now treat automated AI R&D as a competitive advantage.
Related event: Baidu Unveils ERNIE Evolve, an Automated AI Research System(2 posts)→
More from coding & agent
- Solo dev's infinitely evolving Claude-built Backrooms game hits 35K players in 36 hours — minchoi · 2026-11-09
- RentAHuman lets AI agents hire real people for physical-world tasks via MCP API — conitzer · 2026-11-09
- Jane Street open-sources /socratic-tutor: an agent skill that teaches via questioning, not lecturing — daniel_mac8 · 2026-11-09
- Fly-by-Code: coding agents pilot drones via active visual and physical feedback loops — jbhuang0604 · 2026-11-09
- Unreal Engine 5.9 preview is the first UE release built primarily with agentic workflows — nicolascraske · 2026-11-09
- YC-backed Floworks launches Eniac, an API for verified large-scale web data extraction — ycombinator · 2026-11-09